Dev10x:db

Database Query Planning

Orchestration

This skill follows references/task-orchestration.md patterns. Create a task at invocation, mark completed when done:

REQUIRED: Create a task at invocation. Execute at startup:

1. TaskCreate(subject="Plan database query", activeForm="Planning query")

Mark completed when done: TaskUpdate(taskId, status="completed")

Schema-first query construction for safe, accurate database queries.

Query Planning Workflow

Before writing any SQL:

1. Load schema context — search for schema memory files and project-specific db skills (see Context Discovery below)
2. Verify table and column names — never guess; cross-reference the schema files or run an information_schema query
3. Write the query — use verified names only
4. Execute via engine-specific skill — delegate to the appropriate execution layer (e.g., Dev10x:db-psql for PostgreSQL)

Context Discovery Convention

Generic skills discover database context by searching these resources (in order, first match wins per resource type):

Schema files

1. ~/.claude/memory/Dev10x/db-*-schema.md — global schema docs (preferred, see references/config-resolution.md)

Database configuration

Handled by Dev10x:db-psql — see its SKILL.md for the full search order. Key locations:

1. $DB_CONFIG environment variable (explicit override)
2. Plugin skill directory (skills/db-psql/databases.yaml)
3. ~/.claude/memory/Dev10x/databases.yaml — global, user-level config
4. Sibling plugin skills (skills/*/databases.yaml)
5. User skill directories (~/.claude/skills/*/databases.yaml)

Additional context

• Project db skills: any skill matching *:db pattern (e.g., tt:db, acme:db) — project-specific aliases, common queries, and DSN configuration
• CLAUDE.md database sections: database connection info or references in project instructions

Always load available context before constructing queries.

Safety Rules

1. Read-only queries only — SELECT, WITH (CTEs), EXPLAIN, and SHOW are allowed. Never run INSERT, UPDATE, DELETE, DROP, ALTER, CREATE, TRUNCATE, or other write operations through automated tools
2. Write queries: print the raw SQL and tell the user to run it manually. Never attempt workarounds or bypass safety checks
3. Use timeouts — all queries should have statement timeouts to prevent runaway execution
4. Limit large result sets — use LIMIT on exploratory queries; never SELECT * on production tables without a WHERE clause

Anti-Patterns

Anti-Pattern	What to Do Instead
Guess table/column names	Load schema files first, or query information_schema.tables / information_schema.columns
Retry with random name variations	Stop and look up the correct name in schema docs
SELECT * on large tables	Use specific columns and LIMIT
Inline DSN strings in prompts	Use the engine-specific skill's wrapper script
Skip schema check for "simple" queries	Always verify — even common tables have gotchas

Schema Memory File Format

Schema files should follow this template. Place them at ~/.claude/memory/Dev10x/db-<name>-schema.md (preferred global path per references/config-resolution.md):

# <Project> Database Schema — <database-name>

## <Domain Group>

| Table | Key Columns | Notes |
|-------|-------------|-------|
| `table_name` | id, name, status, foreign_key_id | Brief purpose |

## Common Gotchas

- Table `actual_name` (NOT `guessed_name`)
- Money fields store JSON: `field::jsonb->>'amount'`

## Common Queries

### Find X by Y
\```sql
SELECT ... FROM ... WHERE ...
\```

Integration

This skill provides query planning guidance. Actual query execution is handled by engine-specific skills:

• Dev10x:db-psql — PostgreSQL via psql wrapper